Jocelyne Bachevalier
Introduction
The medial temporal lobe (MTL), which includes the amygdala, hippo- campal formation (dentate gyrus, CA fields and subicular complex) and a set of cortical areas (i.e. entorhinal, perirhinal and parahippocampal areas TH and TF) on the parahippocampal gyrus, has long been known to be implicated in memory processes and regulation of emotional behaviours. In experimental work with animals, the development of new surgical tools that have allowed more selective lesions of the amygdala and hippocam- pus (e.g. sparing adjacent temporal cortical areas), together with neuro- psychological studies of the functions of the various medial temporal cortical areas have helped to clarify the contributions of each of these medial temporal lobe components to cognitive functions (Baxter & Murray, 2000; Brown & Aggleton, 2001; Eichenbaum, 2001; Lavenex & Amaral, 2000; Mishkin et al., 1997; Murray & Bussey, 1999; O’Reilly & Rudy, 2001; Yonelinas, 2002). Although controversies still exist in the field, the general picture that emerges can be summarized as follows. The perirhinal cortex, which receives perceptual information about objects, mediates item-specific memory, as well as learning of stimulus–stimulus, cross-modal and stimulus–reward associations. In turn, areas TH and TF, which receive more extensive spatial information about objects, are involved in spatial memory. Thus, these temporal cortical areas are involved in the storage and retrieval of stimulus representations, and are viewed as storing information or knowledge independently of the context in which they are learned (fact or semantic memory). By comparison, the hippocampal formation is needed to acquire, store and recollect inter- item relations and their context and supports recollection of specific events (relational, configural and/or episodic memory). Finally, although the amygdala is not involved in stimulus memory and in learning stimulus– stimulus, cross-modal and stimulus–reward associations per se, it plays a critical role in a specific class of stimulus–reward associations, namely the associations of a stimulus with the specific or current value of a
reinforcer. In addition, the amygdala acts as an interface between stim- ulus, both animate and inanimate, and emotional responses. Thus, selec- tive amygdala lesions result in emotional changes (e.g. reduced fear and aggressiveness, increased submission, and excessive manual and oral exploration), whilst sparing behavioural and physiological components of anxious temperament (Kalin et al., 2001, 2004; Meunier et al., 1999). Further, the magnitude of these emotional changes is generally greater in monkeys with aspiration lesions of the amygdala that include portions of the medial temporal cortex in addition to the amygdala, suggesting a role for these cortical areas in emotional behaviour as well. This is supported by recent findings in monkeys showing that selective damage to the ento- and perirhinal cortex is accompanied by emotional changes that are in fact the opposite of those found after neurotoxic lesions of the amygdala, namely enhanced fearful reactions (Meunier & Bachevalier, 2002; Meunier, Cirilli & Bachevalier, 2006).
Much less emphasis has been placed, however, on the development of cognitive functions mediated by the medial temporal structures in pri- mates. This chapter will begin with a brief review of the maturation of MTL structures in primates. Current knowledge on the functional devel- opment of MTL structures will then be summarized, relying mostly on developmental neuropsychological studies in nonhuman primates as well as in humans when applicable. The possible relevance of these experi- mental findings for autism will be discussed in a final section.
Maturation of medial temporal lobe structures in primates
As reviewed recently (Alvarado & Bachevalier, 2000; Bachevalier & Vargha-Khadem, 2005; Machado & Bachevalier, 2003a), most of the neurogenesis within the hippocampus occurs before birth in primates. Nevertheless, neurogenesis in the dentate gyrus continues throughout gestation, and even postnatally until the fourth and sixth postnatal months. Yet, despite this early neurogenesis, hippocampal neurons undergo significant postnatal changes, suggesting prolonged morpholog- ical maturation during the first postnatal year in monkeys and up to 5–6 years of age in humans. This postnatal development of the hippocampus has also been shown by an increase in hippocampal volume as well as changes in the ratio of grey to white matter in neuroimaging studies in infant monkeys and humans. Conversely, morphological maturation of the allocortical areas, such as entorhinal and perirhinal cortex, precedes that of the hippocampus, and at birth these cortical areas can be clearly identified cytoarchitectonically and display adult-like chemoanatomical
characteristics. Thus, these cortical areas may be available to support some memory processes (e.g. fact or semantic memory) early in life, whereas memory processes mediated by the hippocampus (relational or episodic memory) may mature progressively over a longer period of time. Finally, neurogenesis in the amygdala occurs early in gestation and pro- ceeds in a smooth dorsal-to-ventral wave in both nonhuman primates and humans. Also, most afferent and efferent connections of amygdala neu- rons are already established by the time of birth, and by the second postnatal week in primates, amygdala efferent connections largely resem- ble those seen in adult monkeys. However, there is still some immaturity in the pattern of afferent cortical projections in the first few postnatal months, suggesting that refinement of cortico-amygdalar projections provides infant monkeys with increasingly detailed information about complex socioemotional stimuli as they mature. This protracted refine- ment in the sensory inputs to the amygdala is accompanied by progressive myelination of the amygdala until the fourth postnatal week, indicating that the influence of the amygdala on other neural systems increases slowly in infancy.
These data suggest that the medial temporal cortical areas and the structural elements and synaptic connections within the hippocampus necessary for memory formation are present in newborn primates, although the modifications of hippocampal circuits from birth to adult- hood provide a basis for memory processes to continue to mature during development. Similarly, although structural elements within the amyg- dala necessary for regulation of emotions are present at birth, postnatal changes in its connectivity with other neural systems could support further maturation of a modulatory role of the amygdala in emotional responses to environmental stimuli.
Given the early development of MTL structures, it could be argued that any insult to this neural system in the perinatal period may result in profound deficits in memory processes and emotional responses. Conversely, given the significant sparing of functions known to occur after early brain damage, it is possible that any early insult to this neural system may result in significant functional compensation. The data reviewed below tend to support the former, rather than the latter, proposal.
Neonatal damage to the medial temporal lobe
My colleagues and I began our exploration of the long-term behavioural and cognitive consequences of neonatal MTL damage by characterizing the memory impairment and socioemotional changes of newborn
monkeys that had received extensive MTL lesions (amygdala, hippo- campal formation and adjacent cortical areas) in the first weeks of life. In these early studies (see review Bachevalier & Ma´lkova´, 2000), we found that neonatal MTL damage resulted in a profound anterograde amnesia, closely resembling the anterograde amnesia that follows similar MTL damage in adult monkeys and humans (Mishkin et al., 1982; Zola- Morgan & Squire, 1985). Thus, unlike age-matched controls, the oper- ated infant monkeys showed impaired performance on item-specific recognition tasks (visual paired comparison and delayed nonmatching- to-sample), as well as in tasks measuring tactual and spatial location recognition. By contrast, they displayed normal immediate memory abil- ities as well as normal performance on tasks of procedural learning, such as concurrent discrimination learning tasks.
Altogether these earlier findings indicated that MTL structures are required for some types of learning and memory processes (fact and episodic memory) but not others (short-term memory, procedural mem- ory or skill learning). This early MTL damage was also accompanied by profound changes in emotionality and social interactions that were greater in magnitude than those found after MTL lesions incurred in adulthood. They were also accompanied by the presence of stereotyped behaviours (Bachevalier, Ma´lkova´ & Mishkin, 2001; Ma´lkova´ et al., 1997).
These behavioural changes led us to suggest that any neural reorgan- ization that occurs after early damage to the MTL structures may be more harmful than beneficial, perhaps because it disrupts the functioning of late-maturing neural systems. This view is now supported by recent data on the same neonatally MTL-operated monkeys and their age-matched controls demonstrating that this early damage interferes with the func- tions of neural systems remote from the site of injury, namely the dorso- lateral prefrontal cortex (Bertolino et al., 1997; Chlan-Fourney et al., 2000, 2003) and the anatomically related portion of the neostriatum (Heinz et al., 2000; Saunders et al., 1998). In addition, the neonatally operated monkeys displayed a dysregulation of prefrontal–striatal dopa- mine transmission that could provide an explanation for the presence of ritualistic and stereotyped behaviours in these animals. Thus, the lack of functional inputs from MTL structures could prevent the prefrontal cortex from undergoing proper neuronal development. However, it is not known at the present time whether these early MTL lesions affect other neural systems; nor is it known which MTL structures are respon- sible for such a prefrontal immaturity.
To learn more about the underlying neural substrate of the behavioural disorders resulting from the early MTL damage, we investigated whether
or not the complex behavioural syndrome described above might be fractionated by damaging specific MTL components. We began this set of studies by producing infant monkeys with damage to the amygdala and hippocampal formation, which were behaviourally tested until adulthood using the same procedures as those used in the previous studies. However, it is important to note that, since the neonatal lesions were produced by aspiration procedures, amygdala damage also included damage to the ento- and perirhinal cortex, and hippocampal damage also included areas TH and TF (for further information on the extent of these neonatal lesions see Bachevalier, Beauregard & Alvarado, 1999). In a preliminary report (Bachevalier, 1994), it was concluded that neo- natal amygdala damage yielded mild deficits in memory function and a pattern of socioemotional deficits almost identical to that found after complete MTL lesions, although the magnitude of the deficits was less. By contrast, early hippocampal damage largely spared memory and resulted in reduced social interactions and locomotor stereotypies only when the monkeys reached adulthood (Beauregard, Ma´lkova´ & Bachevalier, 1995).
Further investigations of these operated animals and their unoperated controls as well as more recent developmental studies of either selective neurotoxic damage to the amygdala and hippocampal formation avoiding the adjacent cortices, or direct damage to the entorhinal and perirhinal cortex avoiding the two subcortical structures, allow more specific char- acterization of the memory and behavioural deficits that follow neonatal damage to different MTL structures. Findings from these recent studies are presented next, subdivided into findings relating first to the amygdala, then to the hippocampal formation, and finally to the temporal cortical areas.
Neonatal damage to the amygdala
Effects on memory
Monkeys with neonatal aspiration lesions of the amygdala (Bachevalier, Beauregard & Alvarado, 1999) showed a mild and transient deficit in procedural learning that was in fact similar to that found earlier in monkeys with neonatal MTL damage. In addition, this early damage yielded significant recognition memory impairment (as assessed using delayed nonmatching-to-sample – DNMS) that persisted when the neo- natally operated animals were re-tested on the same recognition task upon reaching adulthood. As reviewed above, the role of the amygdala in learning stimulus–reward associations such as concurrent discriminations,
as well as in recognition memory as measured by the DNMS, has been recently called into question by findings in adult monkeys indicating that most of the memory impairment resulting from an aspiration lesion of the amygdala appears to be due instead to direct or indirect damage to the ento- and perirhinal cortices (Baxter & Murray, 2000). Thus, the mild and transient impairment in concurrent discrimination learning, and the moderate and long-lasting impairment in the DNMS seen after neonatal amygdaloid lesions and reported by Bachevalier and colleagues (1999), could have resulted from additional damage to cortical areas surrounding the amygdala. This suggestion is supported by recent evidence showing severe recognition memory impairment following direct damage to the ento- and perirhinal cortex in infant monkeys (Ma´lkova´ et al., 1998).
Recent findings on the same neonatally amygdalectomized animals (Alvarado, Wright & Bachevalier, 2002) demonstrated normal abilities on tasks sensitive to selective neurotoxic hippocampal lesions, such as a transverse patterning task (measuring the ability to learn the relationship between three different objects) and a spatial locations recognition task. In addition, our current investigations of the effects of neonatal neurotoxic lesions of the amygdala, which spare the adjacent cortical areas, demon- strate that these early lesions affect neither the learning of stimulus– reward associations, as measured by an object discrimination reversal task, nor the ability to recognize objects over extended delays, as meas- ured by visual paired comparison (VPC). Thus, altogether the data suggest that damage to the amygdala in early infancy, like the same damage in adulthood, does not affect stimulus memory and stimulus–reward associations. However, given that in adult monkeys the amygdala plays a critical role in a particular class of stimulus–reward associations, namely the association of a stimulus with its intrinsic reward value (whether it is pleasant or not), it will be of particular interest to investigate whether the same deficit will be evident in the animals with selective neonatal amygdala damage.
Effects on socioemotional behaviour
Given that the primate amygdala develops early in infancy and plays a critical role in the regulation of emotional responses and social cognition (see for review Adolphs 2001; Bachevalier & Meunier, 2005), it is not surprising to observe that neonatal amygdala lesions also alter emotional responses and social behaviour. Earlier studies (Thompson, Schwartzbaum & Harlow, 1969; Thompson, 1981) have shown that infant monkeys with aspiration amygdala lesions displayed more fear responses during social encounters than did control monkeys with whom they were paired, and
these fear responses became more profound whenever control animals became more active. By contrast, responses to novel objects in the absence of other monkeys revealed an opposite pattern of results, with operated monkeys displaying fewer fear responses than controls. These results sug- gest that the neonatally operated animals have difficulty in evaluating potentially threatening social situations, although they seem uninhibited when confronted with novel objects. When observed in dyadic social inter- actions as they reached adulthood, monkeys with neonatal aspiration lesions of the amygdala, like those with similar lesions in adulthood, showed transient hyperactivity and were subordinate to control animals during social interactions, suggesting that the amygdala lesions may have affected the normal development of aggressive responses (Bachevalier, 1994; Thompson, 1981; Thompson, Bergland & Towfighi, 1977).
Interestingly, recent investigations of social interactions in monkeys with neonatal neurotoxic amygdala lesions and age-matched controls (Bauman et al., 2004; Prather et al., 2001) have demonstrated that the neonatally operated monkeys retained their ability to produce all affili- ative and social gestures, although they expressed more fear, reduced amount of physical contacts with partners and more submissive responses at both 6 and 9 months. Thus, although the social changes following selective neonatal amygdala lesions were much less severe in magnitude than those found when the amygdala damage was associated with cortical damage (as in the case of aspiration lesions), it still remains to be seen whether greater emotional and social changes will emerge as the operated animals from these more recent reports reach maturity.
Summary
Neonatal damage to the amygdala does not alter object recognition memory or stimulus–reward associations, but severely impairs the regu- lation of emotional responses to animate and inanimate stimuli. In addi- tion, although neonatal aspiration lesions of the amygdala that included portions of the medial temporal cortical areas resulted in enhanced social fear and alteration in social interactions that became more severe as the animals reached adulthood, amygdala lesions that spared the cortical areas yielded similar enhanced social fear but milder changes in social interactions. Thus, the data suggest that the behavioural changes seen after neonatal amygdala lesions are exacerbated by additional damage to the adjacent cortical areas. However, future investigations are required to directly assess whether the different behavioural outcomes between these developmental studies reflect variations in lesion extent, rearing condi- tions, timing of social testing, or a combination of these factors.
Neonatal damage to the hippocampal formation
Effects on memory
Monkeys with neonatal aspiration lesions of the hippocampal formation showed no deficits in procedural learning or item-specific recognition memory (Bachevalier, Beauregard & Alvarado, 1999). This lack of mem- ory impairment, however, does not imply that neonatal hippocampal damage spares all memory processes. Indeed, these same neonatally operated animals showed severe impairments on three memory tasks that have been shown to be sensitive to neurotoxic hippocampal lesions in adult monkeys (Alvarado & Bachevalier, 2005; Nemanic, Alvarado & Bachevalier, 2004): recognition memory (VPC task) when long delays are used (Pascalis & Bachevalier, 1999), transverse patterning problems, and recognition of the locations of rewards on a testing board (Alvarado, Wright & Bachevalier, 2002). Because these memory tasks tax recognition- based incidental memory (VPC) and relational memory based on con- junctive representations linking an item with other items (transverse patterning) or with spatial information (spatial DNMS), they have also been thought to measure a precursor form of episodic memory processes in animals (Eichenbaum, 2001). Thus, monkeys with neonatal hippo- campal lesions demonstrated a specific loss in relational (or episodic) memory processes, sparing familiarity-based object recognition as meas- ured by object DNMS (see Nemanic, Alvarado & Bachevalier, 2004 for a discussion of the different declarative memory processes taxed by the VPC and DNMS).
These data on monkeys with neonatal hippocampal lesions parallel those found in children with developmental amnesia who suffer from pathology of the hippocampal formation associated with severe impair- ment in declarative memory processes (see for review Bachevalier & Vargha-Khadem 2005; see also Salmond et al., this volume, Chapter 4). Thus, after early hippocampal pathology, both children and monkeys show a pronounced impairment in context-rich memory processes, including memory for unique events. In individuals with developmental amnesia this impairment in the recall of events and episodes prevails despite the relative preservation of semantic memory and recognition memory (Vargha-Khadem et al., 1997). Furthermore, in at least one individual, neonatal hippocampal pathology appears to selectively impair recollection-based as compared with familiarity-based recognition abil- ities. Whether the deficit in monkeys with neonatal hippocampal lesions in incidental item recognition (VPC), particularly at long delays, is due to the same distinction between familiarity-based and recollection-based
recognition remains to be determined. In the same way, it has yet to be demonstrated whether individuals with developmental amnesia will be impaired in recognition of incidental unique items, especially at long delays, although this is true for adults with acquired damage to the hippocampus (Pascalis et al., 2004).
Effects on socioemotional behaviours
While there is little evidence of behavioural changes following hippo- campal lesions in adult monkeys, a decreased fear towards a human observer has previously been reported (Butter, Mishkin & Mirsky, 1968). Recent studies from our laboratory have shown that selective neurotoxic hippocampal lesions in the adult monkeys result in increased production of tense or anxious signals and activity (hyperlocomotion), as